Natural immunoglobulins have been known for many years, as have the various fragments thereof, such as the Fab, (Fabxe2x80x2)2 and Fc fragments, which can be derived by enzymatic cleavage. Natural immunoglobulins comprise generally a Y-shaped molecule having an antigen-binding site towards the free end of each upper arm. The remainder of the structure, and particularly the stem of the Y, mediates the effector functions associated with immunoglobulins.
Specifically, immunoglobulin molecules are comprised of two heavy (H) and two light (L) polypeptide chains, held together by disulfide bonds. Each chain of an immunoglobulin chain is divided into regions or domains, each being approximately 110 amino acids. The light chain has two such domains while the heavy chain has four domains. The amino acid sequence of the amino-terminal domain of each polypeptide chain is highly variable (V region), while the sequences of the remaining domains are conserved or constant (C regions). A light chain is therefore composed of one variable (VL) and one constant domain (CL) while a heavy chain contains one variable (VH) and three constant domains (CH1, CH2 and CH3). An arm of the Y-shaped molecule consists of a light chain (V+CL) and the variable domain (VH) and one constant domain (CH1) of a heavy chain. The tail of the Y is composed of the remaining heavy chain constant domains (CH2+CH3). The C-terminal ends of the heavy chains associate to form the Fc portion. Within each variable region are three hypervariable regions. These hypervariable regions are also described as the complementarity determining regions (CDRs) because of their importance in binding of antigen. The four more conserved regions of the variable domains are described as the framework regions (FRs). Each domain of an immunoglobulin consists of two beta-sheets held together by a disulfide bridge, with their hydrophobic faces packed together. The individual beta strands are linked together by loops. The overall appearance can be described as a beta barrel having loops at the ends. The CDRs form the loops at one end of the beta barrel of the variable region.
Natural immunoglobulins have been used in assay, diagnosis and, to a more limited extent, therapy. However, such uses, especially in therapy, have been hindered by the polyclonal nature of natural immunoglobulins. A significant step towards the realization of the potential of immunoglobulins as therapeutic agents was the discovery of techniques for the preparation of monoclonal antibodies (MAbs) of defined specificity, Kohler et al., 1975 [1]. However, most MAbs are produced by fusions of rodent (i.e., mouse, rat) spleen cells with rodent myeloma cells. They are therefore essentially rodent proteins.
By 1990, over 100 murine monoclonal antibodies were in clinical trials, particularly in the U.S. and especially for application in the treatment of cancer. However, by this time it was recognized that rejection of murine monoclonal antibodies by the undesirable immune response in humans termed the HAMA (Human Anti-Mouse Antibody) response was a severe limitation, especially for the treatment of chronic disease. Therefore, the use of rodent MAbs as therapeutic agents in humans is inherently limited by the fact that the human subject will mount an immunological response to the MAb and either remove the MAb entirely or at least reduce its effectiveness. In practice MAbs of rodent origin may not be used in a patient for more than one or a few treatments as a HAMA response soon develops rendering the MAb ineffective as well as giving rise to undesirable reactions. In fact, a HAMA response has been observed in the majority of patients following a single injection of mouse antibody, (Schroff et al., 1985 [2]). A solution to the problem of HAMA is to administer immunologically compatible human monoclonal antibodies. However, the technology for development of human monoclonal antibodies has lagged well behind that of murine antibodies (Borrebaeck et al., 1990 [3]) such that very few human antibodies have proved useful for clinical study.
Proposals have therefore been made for making non-human MAbs less antigenic in humans. Such techniques can be generically termed xe2x80x9chumanizationxe2x80x9d techniques. These techniques generally involve the use of recombinant DNA technology to manipulate DNA sequences encoding the polypeptide chains of the antibody molecule. The use of recombinant DNA technology to clone antibody genes has provided an alternative whereby a murine monoclonal antibody can be converted to a predominantly human-form (i.e., humanized) with the same antigen binding properties (Riechmann et al., 1988 [4]). Generally, the goal of the humanizing technology is to develop humanized antibodies with very little or virtually no murine component apart from the CDRs (see, e.g., Tempest et al., 1991 [5]) so as to reduce or eliminate their immunogenicity in humans.
Early methods for humanizing MAbs involved production of chimeric antibodies in which an antigen binding site comprising the complete variable domains of one antibody is linked to constant domains derived from another antibody. Methods for carrying out such chimerization procedures have been described, for example, in EP 120694 [6], EP 125023 [7], and WO 86/01533 [8]. Generally disclosed are processes for preparing antibody molecules having the variable domains from a non-human MAb such as a mouse MAb and the constant domains from a human immunoglobulin. Such chimeric antibodies are not truly humanized because they still contain a significant proportion of non-human amino acid sequence, i.e., the complete non-human variable domains, and thus may still elicit some HAMA response, particularly if administered over a prolonged period, Begent et al., 1990 [9]. In addition, it is believed that these methods in some cases (e.g., EP 120694 [6]; EP 125023 [7] and U.S. Pat. No. 4,816,567 [10]) did not lead to the expression of any significant quantities of Ig polypeptide chains, nor the production of Ig activity without in vitro solubilization and chain reconstitution, nor to the secretion and assembly of the chains into the desired chimeric recombinant antibodies. These same problems may be noted for the initial production of non-chimeric recombinant antibodies (e.g., U.S. Pat. No. 4,816,397 [11]).
Following the early methods for the preparation of chimeric antibodies, a new approach was described in EP 0239400 [12] whereby antibodies are altered by substitution of their complementarity determining regions (CDRs) for one species with those from another. This process may be used, for example, to substitute the CDRs from human heavy and light chain Ig variable region domains with alternative CDRs from murine variable region domains. These altered Ig variable regions may subsequently be combined with human Ig constant regions to created antibodies which are totally human in composition except for the substituted murine CDRs. Such murine CDR-substituted antibodies would be predicted to be less likely to elicit a considerably reduced immune response in humans compared to chimeric antibodies because they contain considerably less murine components.
The process for humanizing monoclonal antibodies via CDR grafting has been termed xe2x80x9creshapingxe2x80x9d. (Riechmann et al., 1988 [4]; Verhoeyen et al., 1988 [13]). Typically, complementarity determining regions (CDRs) of a murine antibody are transplanted onto the corresponding regions in a human antibody, since it is the CDRs (three in antibody heavy chains, three in light chains) that are the regions of the mouse antibody which bind to a specific antigen. Transplantation of CDRs is achieved by genetic engineering whereby CDR DNA sequences are determined by cloning of murine heavy and light chain variable (V) region gene segments, and are then transferred to corresponding human V regions by site-directed mutagenesis. In the final stage of the process, human constant region gene segments of the desired isotype (usually gamma 1 for CH and kappa for CL) are added and the humanized heavy and light chain genes are coexpressed in mammalian cells to produce soluble humanized antibody.
The transfer of these CDRs to a human antibody confers on this antibody the antigen binding properties of the original murine antibody. The six CDRs in the murine antibody are mounted structurally on a V region xe2x80x9cframeworkxe2x80x9d region. The reason that CDR-grafting is successful is that framework regions between mouse and human antibodies may have very similar 3-D structures with similar points of attachment for CDRs, such that CDRs can be interchanged. Nonetheless, certain amino acids within framework regions are thought to interact with CDRs and to influence overall antigen binding affinity. The direct transfer of CDRs from a murine antibody to produce a recombinant humanized antibody without any modifications of the human V region frameworks often results in a partial or complete loss of binding affinity.
In Riechmann et al., 1988 [4] and WO 89/07454 [14], it was found that transfer of the CDR regions alone (as defined by Kabat et al., 1991 [15] and Wu et al., 1970 [16]) was not sufficient to provide satisfactory antigen binding activity in the CDR-grafted product. Riechmann et al. 1988 [4] found that it was necessary to convert a serine residue at position 27 of the human sequence to the corresponding rat phenylalanine residue to obtain a CDR-grafted product having satisfactory antigen binding activity. This residue at position 27 of the heavy chain is within the structural loop adjacent to CDR1. A further construct which additionally contained a human serine to rat tyrosine change at position 30 of the heavy chain did not have a significantly altered binding activity over the humanized antibody with the serine to phenylalanine change at position 27 alone. These results indicate that changes to residues of the human sequence outside the CDR regions, for example, in the loop adjacent to CDR1, may be necessary to obtain effective antigen binding activity for CDR-grafted antibodies which recognize more complex antigens. Even so, the binding affinity of the best CDR-grafted antibodies obtained was still significantly less than the original MAb.
More recently, Queen et al., 1989 [17] and WO 90/07861 [18] have described the preparation of a humanized antibody that binds to the interleukin 2 receptor, by combining the CDRs of a murine MAb (anti-Tac) with human immunoglobulin framework and constant regions. They have demonstrated one solution to the problem of the loss of binding affinity that often results from direct CDR transfer without any modifications of the human V region framework residues; their solution involves two key steps. First, the human V framework regions are chosen by computer analysts for optimal protein sequence homology to the V region framework of the original murine antibody, in this case, the anti-Tac MAb. In the second step, the tertiary structure of the murine V region is modelled by computer in order to visualize framework amino acid residues which are likely to interact with the murine CDRs and these murine amino acid residues are then superimposed on the homologous human framework. Their approach of employing homologous human frameworks with putative murine contact residues resulted in humanized antibodies with similar binding affinities to the original murine antibody with respect to antibodies specific for the interleukin 2 receptor (Queen et al., 1989 [17]) and also for antibodies specific for herpes simplex virus (HSV) (Co. et al., 1991 [19]). However, the reintroduction of murine residues into human frameworks (at least 9 for anti-interleukin 2 receptor antibodies, at least 9 and 7 for each of two anti-HSV antibodies) may increase the prospect of HAMA response to the framework region in the humanized antibody. Bruggemann et al., 1989 [20] have demonstrated that human V region frameworks are recognized as foreign in mouse, and so, conversely, murine modified human frameworks might give rise to an immune reaction in humans.
According to the above described two step approach in WO 90/07861 [18], Queen et al. outlined four criteria for designing humanized immunoglobulins. The first criterion is to use as the human acceptor the framework from a particular human immunoglobulin that is usually homologous to the non-human donor immunoglobulin to be humanized, or to use a consensus framework from many human antibodies. The second criterion is to use the donor amino acid rather than the acceptor if the human acceptor residue is unusual and the donor residue is typical for human sequences at a specific residue of the framework. The third criterion is to use the donor framework amino acid residue rather than the acceptor at positions immediately adjacent to the CDRs. The fourth criterion is to use the donor amino acid residue at framework positions at which the amino acid is predicted to have a side chain atom within about 3 xc3x85 of the CDRs in a three-dimensional immunoglobulin model and to be capable of interacting with the antigen or with the CDRs of the humanized immunoglobulin. It is proposed that criteria two, three or four may be applied in addition or alternatively to criterion one, or each criteria may be applied singly or in any combination.
In addition, WO 90/07861 [18] details the preparation of a single CDR-grafted humanized antibody, a humanized antibody specificity for the p55 Tac protein of the IL-2 receptor, by employing the combination of all four criteria, as above, in designing this humanized antibody. The variable region frameworks of the human antibody EU (see, Kabat et al., 1991 [15]) were used as acceptor. In the resultant humanized antibody, the donor CDRs were as defined by Kabat et al., 1991 [15] and Wu et al., 1970 [16] and, in addition, the mouse donor residues were used in place of the human acceptor residues, at positions 27, 30, 48, 66, 67, 89, 91, 94, 103, 104, 105 and 107 in heavy chain and at positions 48, 60 and 63 in the light chain, of the variable region frameworks. The humanized anti-Tac antibody obtained was reported to have an affinity for p55 of 3xc3x97109 Mxe2x88x921, about one-third of that of the murine MAb.
Several other groups have demonstrated that Queen et al.""s approach of first choosing homologous frameworks followed by reintroduction of mouse residues may not be necessary to achieve humanized antibodies with similar binding affinities to the original mouse antibodies (Riechmann et al., 1988 [4]; Tempest et al., 1991 [5]; Verhoeyen, et al. 1991 [21]). Moreover, these groups have used a different approach and have demonstrated that it is possible to utilize, as standard, the V region frameworks derived from NEWM and REI heavy and light chains respectively for CDR-grafting without radical introduction of mouse residues. However, the determination of which mouse residues should be introduced to produce antibodies with binding efficiencies similar to the original murine MAb can be difficult to predict, being largely empirical and not taught by available prior art. In the case of the humanized CAMPATH-IH antibody, the substitution of a phenylalanine for a serine residue at position 27 was the only substitution required to achieve a binding efficiency similar to that of the original murine antibody (Riechmann, et al., 1988 [4]; WO92/04381 [22]). In the case of a humanized (reshaped) antibody specific for respiratory syncytial virus (RSV) for the inhibition of RSV infection in vivo, substitution of a block of 3 residues adjacent to CDR3 in the CDR-grafted NEWM heavy chain was required to produce biological activity equivalent to the original mouse antibody (Tempest et al., 1991 [5]; WO92/04381 [22]). The reshaped antibody in which only the mouse CDRs were transferred to the human framework showed poor binding for RSV. An advantage of using the Tempest et al., 1991 [5] approach to construct NEWM and REI based humanized antibodies is that the 3-dimensional structures of NEWM and REI variable regions are known from x-ray crystallography and thus specific interactions between CDRs and V region framework residues can be modelled.
Regardless of the approach taken, the examples of the initial humanized antibodies prepared to date have shown that it is not a straightforward process to obtain humanized antibodies with the characteristics, in particular, the binding affinity, as well as other desirable properties, of the original murine MAb from which the humanized antibody is derived. Regardless of the approach to CDR grafting taken, it is often not sufficient merely to graft the CDRs from a donor Ig onto the framework regions of an acceptor Ig (see, e.g., Tempest et al., 1991 [5], Riechmann et al., 1988 [4], etc., cited herein). In a number of cases, it appears to be critical to alter residues in the framework regions of the acceptor antibody in order to obtain binding activity. However, even acknowledging that such framework changes may be necessary, it is not possible to predict, on the basis of the available prior art, which, if any, framework residues will need to be altered to obtain functional humanized recombinant antibodies of the desired specificity. Results thus far indicate that changes necessary to preserve specificity and/or affinity are for the most part unique to a given antibody and cannot be predicted based on the humanization of a different antibody.
In particular, the sets of residues in the framework region which are herein disclosed as being of critical importance to the activity of the recombinant humanized anti-VLA-4 antibodies constructed in accordance with the teachings of the present invention do not generally coincide with residues previously identified as critical to the activity of other humanized antibodies and were not discovered based on the prior art.
To date, humanized recombinant antibodies have been developed mainly for therapeutic application in acute disease situations (Tempest, et al., 1991 [5]) or for diagnostic imaging (Verhoeyen, et al., 1991 [21]). Recently, clinical studies have begun with at least two humanized antibodies with NEWM and REI V region frameworks, CAMPATH-IH (Riechmann et al., 1988 [4]) and humanized anti-placental alkaline phosphatase (PLAP) (Verhoeyen et al., 1991 [21]) and these studies have initially indicated the absence of any marked immune reaction to these antibodies. A course of treatment with CAMPATH-IH provided remission for two patients with non-Hodgkin""s lymphoma thus demonstrating efficacy in a chronic disease situation (Hale et al., 1988 [23]). In addition, the lack of immunogenicity of CAMPATH-1H was demonstrated after daily treatment of the two patients for 30 and 43 days. Since good tolerance to humanized antibodies has been initially observed with CAMPATH-IH, treatment with humanized antibody holds promise for the prevention of acute disease and to treatment of diseases with low mortality.
Vascular endothelial cells constitute the lining of blood vessels and normally exhibit a low affinity for circulating leukocytes (Harlan, 1985 [24]). The release of cytokines at sites of inflammation, and in response to immune reactions, causes their activation and results in the increased expression of a host of surface antigens. (Collins et al., 1986 [25]; Pober et al., 1986 [26]; Bevilacqua et al., 1987 [27]; Leeuwenberq et al., 1989 [28]). These include the adhesion proteins ELAM-1, which binds neutrophils (Bevilacqua et al., 1989 [29], ICAM-1 which interacts with all leukocytes (Dustin et al., 1986 [30]; Pober et al. 1986, [26]; Boyd et al., 1988 [31]; Dustin and Springer, 1988 [32]), and VCAM-1 which binds lymphocytes (Osborn et al., 1989 [33]). These cytokine-induced adhesion molecules appear to play an important role in leukocyte recruitment to extravascular tissues.
The integrins are a group of cell-extracellular matrix and cell-cell adhesion receptors exhibiting an alpha-beta heterodimeric structure, with a widespread cell distribution and a high degree of conservation throughout evolution (Hynes, 1987 [34]; Marcantonio and Hynes, 1988 [35]). The integrins have been subdivided into three major subgroups; the xcex22 subfamily of integrins (LFA-1, Mac-1, and p150, 95) is mostly involved in cell-cell interactions within the immune system (Kishimoto et al., 1989 [36]), whereas members of the xcex21 and xcex23 integrin subfamilies predominantly mediate cell attachment to the extracellular matrix (Hynes, 1987 [34]; Ruoslahti, 1988 [37]). In particular, the xcex21 integrin family, also termed VLA proteins, includes at least six receptors that specifically interact with fibronectin, collagen, and/or laminin (Hemler, 1990 [38]). Within the VLA family, VLA-4 is atypical because it is mostly restricted to lymphoid and myeloid cells (Hemler et al., 1987 [39]), and indirect evidence had suggested that it might be involved in various cell-cell interactions (Clayberger et al., 1987 [40]; Takada et al., 1989 [41]; Holtzmann et al., 1989 [42]; Bendarczyk and McIntyre, 1990 [43]). In addition, VLA-4 has been shown to mediate T and B lymphocyte attachment to the heparin II binding fragment of human plasma fibronectin (FN) (Wayner et al., 1989 [44]).
VCAM-1, like ICAM-1, is a member of the immunoglobulin gene superfamily (Osborn et al., 1989 [33]). VCAM-1 and VLA-4 were demonstrated to be a ligand-receptor pair that allows attachment of lymphocytes to activated endothelium by Elices et al., 1990 [45]. Thus, VLA-4 represents a singular example of a xcex21 integrin receptor participating in both cell-cell and cell-extracellular matrix adhesion functions by means of the defined ligands VCAM-1 and FN.
VCAM1 (also known as INCAM-110) was first identified as an adhesion molecule induced on endothelial cells by inflammatory cytokines (TNF and IL-1) and LPS (Rice et al., 1989 [46]; Osborn et al., 1989 [33]). Because VCAM1 binds to cells exhibiting the integrin VLA-4 (xcex14xcex21) including T and B lymphocytes, monocytes, and eosinophils, but not neutrophils, it is thought to participate in recruitment of these cells from the bloodstream to areas of infection and inflammation (Elices et al, 1990 [45]; Osborn, 1990 [33]). The VCAM1/VLA-4 adhesion pathway has been associated with a number of physiological and pathological processes. Although VLA-4 is normally restricted to hematopoietic lineages, it is found on melanoma cell lines, and thus it has been suggested that VCAM1 may participate in metastasis of such tumors (Rice et al., 1989 [46]).
In vivo, VCAM1 is found on areas of arterial endothelium representing early atherosclerotic plaques in a rabbit model system (Cybulsky and Gimbrone, 1991 [47]). VCAM1 is also found on follicular dendritic cells in human lymph nodes (Freedman et al., 1990 [48]). It is also present on bone marrow stromal cells in the mouse (Miyake et al., 1991 [49]), thus VCAM1 appears to play a role in B-cell development.
The major form of VCAM1 in vivo on endothelial cells, has been referred to as VCAM-7D, and has seven Ig homology units or domains; domains 4, 5 and 6 are similar in amino acid sequence to domains 1, 2 and 3, respectively, suggesting an intergenic duplication event in the evolutionary history of the gene (Osborn et al., 1989 [33]; Polte et al. 1990 [50]; Hession et al., 1991 [51]; Osborn and Benjamin, U.S. Ser. No. 07/821,712 filed Sep. 30, 1991, [52]). A 6-domain form (referred to as VCAM-6D herein) is generated by alternative splicing, in which the fourth domain is deleted (Osborn et al., 1989 [33]; Hession et al. 1991 [51], Cybulsky et al., 1991 [47]; Osborn and Benjamin, U.S. Ser. No. 07/821,712 filed Sep. 30, 1991 [52]). The VCAM-6D, was the first sequenced of these alternate forms, however, later in vivo studies showed that the VCAM-7D form was dominant in vivo. The biological significance of the alternate splicing is not known, however as shown by Osborn and Benjamin, U.S. Ser. No. 07/821,712 filed Sep. 30, 1991 [52], VCAM-6D can bind VLA-4-expressing cells and thus clearly has potential functionality in vivo.
The apparent involvement of the VCAM1/VLA-4 adhesion pathway in infection, inflammation and possibly atherosclerosis has led to continuing intensive research to understand the mechanisms of cell-cell adhesion on a molecular level and has led investigators to propose intervention in this adhesion pathway as a treatment for diseases, particularly inflammation (Osborn et al., 1989 [33]). One method of intervention in this pathway could involve the use of anti-VLA-4 antibodies.
Monoclonal antibodies that inhibit VCAM1 binding to VLA-4 are known. For example, anti-VLA-4 MAbs HP2/1 and HP1/3 have been shown to block attachment of VLA-4-expressing Ramos cells to human umbilical vein cells and VCAM1-transfected COS cells (Elices et al., 1990 [45]). Also, anti-VCAM1 antibodies such as the monoclonal antibody 4B9 (Carlos et al., 1990 [53]) have been shown to inhibit adhesion of Ramos (B-cell-like), Jurkat (T-cell-like) and HL60 (granulocyte-like) cells to COS cells transfected to express VCAM-6D and VCAM-7D (Hession et al., 1991 [51]).
The monoclonal antibodies to VLA-4 that have been described to date fall into several categories based on epitope mapping studies (Pulido, et al., 1991 [54]). Importantly one particular group of antibodies, to epitope xe2x80x9cBxe2x80x9d, are effective blockers of all VLA-4-dependent adhesive functions (Pulido et al., 1991, [54]). The preparation of such monoclonal antibodies to epitope B of VLA 4, including, for example the HP1/2 MAb, have been described by Sanchez-Madrid et al., 1986, [55]. Antibodies having similar specificity and having high binding affinities to VLA-4 comparable to that of HP1/2, would be particularly promising candidates for the preparation of humanized recombinant anti-VLA-4 antibodies useful as assay reagents, diagnostics and therapeutics.
As stated above, inflammatory leukocytes are recruited to sites of inflammation by cell adhesion molecules that are expressed on the surface of endothelial cells and which act as receptors for leukocyte surface proteins or protein complexes. In particular, eosinophils have recently been found to participate in three distinct cell adhesion pathways to vascular endothelium, binding to cells expressing intercellular adhesion molecule-1 (ICAM-1), endothelial cell adhesion molecule-1 (ELAM-1), and vascular cell adhesion molecule-1 (VCAM-1) (Weller et al., 1991 [56]; Walsh et al., 1991 [57]; Bochner et al., 1991 [58]; and Dobrina et al., 1991 [59]). That eosinophils express VLA-4 differentiates them from other inflammatory cells such as neutrophils, which bind to ELAM-1 and ICAM-1 but not VCAM-1.
The VLA-4-mediated adhesion pathway has been investigated in an asthma model to examine the possible role of VLA-4 in leukocyte recruitment to inflamed lung tissue (Lobb, U.S. Ser. No. 07/821,768 filed Jan. 13, 1992 [60]). Administering anti-VLA-4 antibody inhibited both the late phase response and airway hyperresponsiveness in allergic sheep. Surprisingly, administration of anti-VLA-4 led to a reduction in the number of both neutrophils and eosinophils in the lung at 4 hours after allergen challenge, even though both cells have alternate adhesion pathways by which they can be recruited to lung tissues. Also surprisingly, inhibition of hyperresponsiveness in the treated sheep was observed which continued to 1 week, even though infiltration of leukocytes, including neutrophils and eosinophils, was not significantly reduced over time.
The VLA-4-mediated adhesion model has also been investigated in a primate model of inflammatory bowel disease (IBD) (Lobb, U.S. Ser. No, 07/835,139 filed Feb. 12, 1992 [61]). The administration of anti-VLA-4 antibody surprisingly and significantly reduced acute inflammation in that model, which is comparable to ulcerative colitis in humans.
More recently, anti-VLA-4 antibodies have been used in methods for the peripheralizing of CD34+ cells, including hematopoietic stem cells as described in Papayannopoulou, U.S. Ser. No. 07/977,702, filed Nov. 13, 1992 [62].
Thus, anti-VLA-4 antibodies having certain epitopic specificities and certain binding affinities may be therapeutically useful in a variety of inflammatory conditions, including asthma and IBD. In particular, humanized recombinant versions of such anti-VLA-4 antibodies, if they could be constructed, might be especially useful for administration in humans. Such humanized antibodies would have the desired potency and specificity, while avoiding or minimizing an immunological response which would render the antibody ineffective and/or give rise to undesirable side effects.
The present invention provides a method of constructing a recombinant anti-VLA-4 antibody molecule. Specifically, recombinant antibodies according to the present invention comprise the antigen binding regions derived from the heavy and/or light chain variable regions of an anti-VLA-4 antibody.
Monoclonal antibodies such as HP1/2 and other anti-VLA-4 antibodies (e.g., Mab HP-2/1, HP2/4, L25, P4C2) capable of recognizing the xcex1 chain of VLA-4 will be useful in the present invention. xe2x80x9cIt is most preferred that the antibodies will recognize epitope B of the VLA4-xcex14 chain (see Pulido et al. J. Biol. Chem., 266(16); 10241-10245 (1991).xe2x80x9d
The present invention provides a method for the construction of humanized recombinant antibody molecule using as a first step CDR grafting or xe2x80x9creshapingxe2x80x9d technology. Specifically, the humanized antibodies according to the present invention have specificity for VLA-4 and have an antigen binding site wherein at least one or more of the complementarity determining regions (CDRs) of the variable domains are derived from a donor non-human anti-VLA-4 antibody, and in which there may or may not have been minimal alteration of the acceptor antibody heavy and/or light variable framework region in order to retain donor antibody binding specificity. Preferably, the antigen binding regions of the CDR-grafted heavy chain variable domain comprise the CDRs corresponding to positions 31-35 (CDR1), 50-65 (CDR2) and 95-102 (CDR3). In a preferred embodiment, the heavy chain further includes non-human residues at framework positions 27-30 (Kabat numbering). The heavy chain can further include non-human residues at framework position 75 (Kabat numbering). The heavy chain can further include non-human residues at framework position(s) 77-79 or 66-67 and 69-71 or 84-85 or 38 and 40 or 24 (Kabat numbering). Preferably, the antigen binding regions of the CDR-grafted light chain variable domain comprise CDRs corresponding to positions 24-34 (CDR1), 50-56 (CDR2) and 89-97 (CDR3). In a preferred embodiment, the light chain further includes non-human residues at framework positions 60 and 67 (Kabat numbering). These residue designations are numbered according to the Kabat numbering (Kabat et al., 1991 [15]). Thus, the residue/position designations do not always correspond directly with the linear numbering of the amino acid residues shown in the sequence listing. In the case of the humanized VK sequence disclosed herein, the Kabat numbering does actually correspond to the linear numbering of amino acid residues shown in the sequence listing. In contrast, in the case of the humanized VH sequences disclosed herein, the Kabat numbering does not correspond to the linear numbering of amino acid residues shown in the sequence listing (e.g., for the humanized VH regions disclosed in the sequence listing, CDR2=50-66, CDR3=99-110).
The invention further provides the recombinant and humanized anti-VLA-4 antibodies which may be detectably labelled.
The invention additionally provides a recombinant DNA molecule capable of expressing the recombinant and humanized anti-VLA-4 antibodies of the present invention. The invention further provides host cells capable of producing the recombinant and humanized anti-VLA-4 antibodies of the present invention.
The invention additionally relates to diagnostic and therapeutic uses for the recombinant and humanized anti-VLA-4 antibodies of the present invention.
The invention further provides a method for treating inflammation resulting from a response of the specific defense system in a mammalian subject, including humans, which comprises providing to a subject in need of such treatment an amount of an anti-inflammatory agent sufficient to suppress the inflammation wherein the anti-inflammatory agent is a recombinant and humanized anti-VLA-4 antibody of the present invention.
The invention further provides a method for treating non-specific inflammation in a mammalian subject, including humans using the recombinant and humanized anti-VLA-4 antibodies.
The invention further concerns the embodiment of the above-described methods wherein the recombinant and humanized anti-VLA-4 antibodies of the present invention are derived from the murine monoclonal antibody HP1/2.
The technology for producing monoclonal antibodies is well known. Briefly, an immortal cell line (typically myeloma cells) is fused to lymphocytes (typically splenocytes) from a mammal immunized with whole cells expressing a given antigen, e.g., VLA-4, and the culture supernatants of the resulting hybridoma cells are screened for antibodies against the antigen (see, generally, Kohler et al., 1975 [1]).
Immunization may be accomplished using standard procedures. The unit dose and immunization regimen depend on the species of mammal immunized, its immune status, the body weight of the mammal, etc. Typically, the immunized mammals are bled and the serum from each blood sample is assayed for particular antibodies using appropriate screening assays. For example, anti-VLA-4 antibodies may be identified by immunoprecipitation of 125I-labeled cell lysates from VLA-4-expressing cells (see, Sanchez-Madrid et al., 1986 [55] and Hemler et al., 1987 [39]). Anti-VLA-4 antibodies may also be identified by flow cytometry, e.g., by measuring fluorescent staining of Ramos cells incubated with an antibody believed to recognize VLA-4 (see, Elices et al., 1990 [45]). The lymphocytes used in the production of hybridoma cells typically are isolated from immunized mammals whose sera have already tested positive for the presence of anti-VLA-4 antibodies using such screening assays.
Typically, the immortal cell line (e.g., a myeloma cell line) is derived from the same mammalian species as the lymphocytes. Preferred immortal cell lines are mouse myeloma cell lines that are sensitive to culture medium containing hypoxanthine, aminopterin and thymidine (xe2x80x9cHAT mediumxe2x80x9d).
Typically, HAT-sensitive mouse myeloma cells are fused to mouse splenocytes using 1500 molecular weight polyethylene glycol (xe2x80x9cPEG 1500xe2x80x9d). Hybridoma cells resulting from the fusion are then selected using HAT medium, which kills unfused and unproductively fused myeloma cells (unfused splenocytes die after several days because they are not transformed). Hybridomas producing a desired antibody are detected by screening the hybridoma culture supernatants. For example, hybridomas prepared to produce anti-VLA-4 antibodies may be screened by testing the hybridoma culture supernatant for secreted antibodies having the ability to bind to a recombinant xcex14-subunit-expressing cell line, such as transfected K-562 cells (see, e.g., Elices et al., 1990 [45]).
To produce anti VLA-4-antibodies, hybridoma cells that tested positive in such screening assays are cultured in a nutrient medium under conditions and for a time sufficient to allow the hybridoma cells to secrete the monoclonal antibodies into the culture medium. Tissue culture techniques and culture media suitable for hybridoma cells are well known. The conditioned hybridoma culture supernatant may be collected and the anti-VLA-4 antibodies optionally further purified by well-known methods.
Alternatively, the desired antibody may be produced by injecting the hybridoma cells into the peritoneal cavity of an unimmunized mouse. The hybridoma cells proliferate in the peritoneal cavity, secreting the antibody which accumulates as ascites fluid. The antibody may be harvested by withdrawing the ascites fluid from the peritoneal cavity with a syringe.
Several anti-VLA-4 monoclonal antibodies have been previously described (see, e.g., Sanchez-Madrid et al., 1986 [55]; Hemler et al., 1987 [39]; Pulido et al., 1991 [54]). HP1/2 for example, is one such murine monoclonal antibody which recognizes VLA-4. VLA-4 acts as a leukocyte receptor for plasma fibronectin and VCAM-1. Other monoclonal antibodies, such as HP2/1, HP2/4, L25 and P4C2, have been described that also recognize VLA-4.
Recombinant antibodies have been constructed and are described herein in which the CDRs of the variable domains of both heavy and light chains were derived from the murine HP1/2 sequence. Preferred starting materials for constructing recombinant humanized antibodies according to the present invention are anti-VLA-4 antibodies, such as HP1/2, that block the interaction of VLA-4 with both VCAM1 and fibronectin. Particularly preferred are those antibodies, such as HP1/2, which in addition, do not cause cell aggregation. Some anti-VLA-4 blocking antibodies have been observed to cause such aggregation. The HP1/2 MAb (Sanchez-Madrid et al., 1986 [55]) is a particularly excellent candidate for humanization since it has an extremely high potency, blocks VLA-4 interaction with both VCAM1 and fibronectin, but does not cause cell aggregation, and has the specificity for epitope B on VLA-4. In the initial experiments, VH and VK DNA were isolated and cloned from an HP1/2-producing hybridoma cell line. The variable domain frameworks and constant domains for humanization were initially derived from human antibody sequences.
The three CDRs that lie on both heavy and light chains are composed of those residues which structural studies have shown to be involved in antigen binding. Theoretically, if the CDRs of the murine HP1/2 antibody were grafted onto human frameworks to form a CDR-grafted variable domain, and this variable domain were attached to human constant domains, the resulting CDR-grafted antibody would essentially be a human antibody with the specificity of murine HP1/2 to bind human VLA-4. Given the highly xe2x80x9chumanxe2x80x9d nature of this antibody, it would be expected to be far less immunogenic than murine HP1/2 when administered to patients.
However, following testing for antigen binding of a CDR-grafted HP1/2 antibody in which only the CDRs were grafted onto the human framework, it was shown that this did not produce a CDR-grafted antibody having reasonable affinity for the VLA-4 antigen. It was therefore decided that additional residues adjacent to some of the CDRs and critical framework residues needed to be substituted from the human to the corresponding murine HP1/2 residues in order to generate an antibody with binding affinity in the range of 10% to 100% of the binding affinity of the murine HP1/2 MAb. Empirically, changes of one or more residues in the framework regions of VH and VK were made to prepare antibodies of the desired specificity and potency, but without making so many changes in the human framework so as to compromise the essentially human nature of the humanized VH and VK region sequences.
Furthermore, VLA-4-binding fragments may be prepared from the recombinant anti-VLA-4 antibodies described herein, such as Fab, Fabxe2x80x2, F(abxe2x80x2)2, and F(v) fragments; heavy chain monomers or dimers; light chain monomers or dimers; and dimers consisting of one heavy chain and one light chain are also contemplated herein. Such antibody fragments may be produced by chemical methods, e.g., by cleaving an intact antibody with a protease, such as pepsin or papain, or via recombinant DNA techniques, e.g., by using host cells transformed with truncated heavy and/or light chain genes. Heavy and light chain monomers may similarly be produced by treating an intact antibody with a reducing agent such as dithiothreitol or xcex2-mercaptoethanol or by using host cells transformed with DNA encoding either the desired heavy chain or light chain or both.